database routing breaks at a restart of InfluxDB

[x70b1]

I setup database routing. Every service (MySQL, RabbitMQ and Jenkins) should use it's own database in InfluxDB. The used config is attached.
This works good so far.

If I restart the influxdb service the routing is broken for 3 to 4 data points (Interval is 10s).
This means that every collected data point is written to every configured database.

• I got 3 to 4 MySQL, RabbitMQ and Jenkins data points in my MySQL database.
• I got 3 to 4 MySQL, RabbitMQ and Jenkins data points in my RabbitMQ database.
• I got 3 to 4 MySQL, RabbitMQ and Jenkins data points in my Jenkins database.

This is not a very critical error but it's causing confusion.
Also I am not sure if this is a problem with Telegraf, because InfluxDB is also involved.

I could not find any logs related to this. There are lines like read: connection reset by peer. But that's what I'm expecting.

Relevant telegraf.conf:

[[outputs.influxdb]]
  urls = ["http://127.0.0.1:8086"]
  username = "telegraf"
  password = "xxx"
  database = "RabbitMq"

  skip_database_creation = true
  database_tag = "influxdb_database"
  exclude_database_tag = true

[[outputs.influxdb]]
  urls = ["http://127.0.0.1:8086"]
  username = "telegraf"
  password = "xxx"
  database = "Jenkins"

  skip_database_creation = true
  database_tag = "influxdb_database"
  exclude_database_tag = true

[[outputs.influxdb]]
  urls = ["http://127.0.0.1:8086"]
  username = "telegraf"
  password = "xxx"
  database = "Mysql"

  skip_database_creation = true
  database_tag = "influxdb_database"
  exclude_database_tag = true

[[inputs.rabbitmq]]
  url = "https://rabbit.example.org"
  username = "xxx"
  password = "xxx"

  [inputs.rabbitmq.tags]
    influxdb_database = "RabbitMq"

[[inputs.jenkins]]
  url = "https://example.com"
  username = "xxx"
  password = "xxx"

  [inputs.jenkins.tags]
    influxdb_database = "Jenkins"

[[inputs.mysql]]
  servers = ["xxx"]

  [inputs.mysql.tags]
    influxdb_database = "Mysql"

System info:

1.12.0 as Docker Container

Steps to reproduce:

1. check the list of measurements in all 3 databases
2. restart InfluxDB
3. check the list of measurements in all 3 databases

Expected behavior:

Every database should contain its own measurements.

Actual behavior:

A small amount of values get the wrong database.

[danielnelson]

@x70b1 Thanks for the report, as a workaround for the bug you can use this alternate style:

 [[outputs.influxdb]]
  urls = ["http://127.0.0.1:8086"]
  username = "telegraf"
  password = "xxx"
  database = "RabbitMq"
  skip_database_creation = true

  tagexclude = ["influxdb_database"]
  [outputs.influxdb.tagpass]
    influxdb_database = ["RabbitMq"]


[[inputs.rabbitmq]]
  url = "https://rabbit.example.org"
  username = "xxx"
  password = "xxx"

  [inputs.rabbitmq.tags]
    influxdb_database = "RabbitMq"

This will actually be more efficient even when the bug is fixed because only the metrics that match tagpass are added to the output's metric buffer. With the original configuration all metrics are added to every output's metric buffer and then written to the database depending on the influxdb_database tag.